Vehicle Detection¶
In [3]:
import numpy as np
from pathlib import Path
import matplotlib.pyplot as plt
import cv2
import time
from sklearn.svm import LinearSVC
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from skimage.feature import hog
from scipy.ndimage.measurements import label
from collections import deque
%matplotlib inline
ROOT_PATH = Path('/home/downloads/carnd-p5-vehicle-detection')
Utilities¶
In [11]:
def read_image(img_path):
if img_path.endswith('.png'):
img = plt.imread(img_path)*255
img = np.clip(img, 0, 255).astype(np.uint8)
else:
img = plt.imread(img_path)
return img
def bin_spatial(img, size=(32, 32)):
features = cv2.resize(img, size).ravel()
return features
def color_hist(img, nbins=32): #bins_range=(0, 256)
# Compute the histogram of the color channels separately
channel1_hist = np.histogram(img[:,:,0], bins=nbins)
channel2_hist = np.histogram(img[:,:,1], bins=nbins)
channel3_hist = np.histogram(img[:,:,2], bins=nbins)
# Concatenate the histograms into a single feature vector
hist_features = np.concatenate((channel1_hist[0], channel2_hist[0], channel3_hist[0]))
# Return the individual histograms, bin_centers and feature vector
return hist_features
def convert_color_rgb(image, cspace='RGB'):
if cspace != 'RGB':
if cspace == 'HSV':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV)
elif cspace == 'LUV':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2LUV)
elif cspace == 'HLS':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2HLS)
elif cspace == 'YUV':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2YUV)
elif cspace == 'YCrCb':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2YCrCb)
elif cspace == 'LAB':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2LAB)
else:
cvt_image = np.copy(image)
return cvt_image
def extract_hog_features(img, orient, pix_per_cell, cell_per_block, feature_vec=True):
hog_features = hog(img, orientations=orient,
pixels_per_cell=(pix_per_cell, pix_per_cell),
cells_per_block=(cell_per_block, cell_per_block),
visualize=False, feature_vector=feature_vec, block_norm='L2-Hys')
return hog_features
def extract_features(image, cspace='RGB', orient=9, pix_per_cell=8,
spatial_size=(32, 32), hist_bins=32,
cell_per_block=2, hog_channel=0, spatial_feat=True,
hist_feat=True, hog_feat=True):
feature_image = convert_color_rgb(image, cspace)
img_features = []
if spatial_feat:
spatial_features = bin_spatial(feature_image, size=spatial_size)
img_features.append(spatial_features)
if hist_feat:
hist_features = color_hist(feature_image, nbins=hist_bins)
img_features.append(hist_features)
if hog_feat:
if hog_channel == 'ALL':
hog_features = []
for ch in range(feature_image.shape[2]):
hog_features_ch = extract_hog_features(image[:, :, ch], orient, pix_per_cell, cell_per_block, feature_vec=True)
hog_features.append(hog_features_ch)
hog_features = np.ravel(hog_features)
else:
hog_features = extract_hog_features(feature_image[:, :, hog_channel], orient,
pix_per_cell, cell_per_block, feature_vec=True)
img_features.append(hog_features)
return np.concatenate(img_features)
def add_heat(heatmap, bbox_list):
# Iterate through list of bboxes
for box in bbox_list:
# Add += 1 for all pixels inside each bbox
# Assuming each "box" takes the form ((x1, y1), (x2, y2))
heatmap[box[0][1]:box[1][1], box[0][0]:box[1][0]] += 1
# Return updated heatmap
return heatmap# Iterate through list of bboxes
def apply_threshold(heatmap, threshold):
# Zero out pixels below the threshold
heatmap[heatmap <= threshold] = 0
# Return thresholded map
return heatmap
def draw_labeled_bboxes(img, labels):
# Iterate through all detected cars
for car_number in range(1, labels[1]+1):
# Find pixels with each car_number label value
nonzero = (labels[0] == car_number).nonzero()
# Identify x and y values of those pixels
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Define a bounding box based on min/max x and y
bbox = ((np.min(nonzerox), np.min(nonzeroy)), (np.max(nonzerox), np.max(nonzeroy)))
# Draw the box on the image
cv2.rectangle(img, bbox[0], bbox[1], (0,0,255), 6)
# Return the image
return img
HOG Features¶
In [12]:
img_car_path = ROOT_PATH/'images/examples/car.png'
img_notcar_path = ROOT_PATH/'images/examples/notcar.png'
img_car = plt.imread(img_car_path.as_posix())
img_notcar = plt.imread(img_notcar_path.as_posix())
print('Image shape', img_car.shape)
plt.imshow(img_car)
plt.show()
plt.imshow(img_notcar)
plt.show()
In [13]:
def vis_hog_features(img, orient, pix_per_cell, cell_per_block, vis=True,
feature_vec=False):
hog_img = None
if vis:
hog_features, hog_img = hog(img, orientations=orient,
pixels_per_cell=(pix_per_cell, pix_per_cell),
cells_per_block=(cell_per_block, cell_per_block),
visualize=vis, feature_vector=feature_vec, block_norm='L2-Hys')
else:
hog_features = hog(img, orientations=orient,
pixels_per_cell=(pix_per_cell, pix_per_cell),
cells_per_block=(cell_per_block, cell_per_block),
visualize=vis, feature_vector=feature_vec, block_norm='L2-Hys')
return hog_features, hog_img
In [14]:
gray = cv2.cvtColor(img_car, cv2.COLOR_RGB2GRAY)
features, hog_img = vis_hog_features(gray, orient=10, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
In [15]:
# Plot the examples
fig = plt.figure(figsize=(12, 9))
plt.subplot(121)
plt.imshow(img_car, cmap='gray')
plt.title('Example Car Image')
plt.subplot(122)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
Out[15]:
Color space¶
In [16]:
test_image_path = ROOT_PATH/'images/test_images/test1.jpg'
test_image = read_image(test_image_path.as_posix())
In [17]:
img = convert_color_rgb(test_image, 'RGB')
fig, axes = plt.subplots(1, 3, figsize=(16, 12))
for i, ax in enumerate(axes):
ax.imshow(img[:, :, i])
In [18]:
img = convert_color_rgb(test_image, 'YUV')
fig, axes = plt.subplots(1, 3, figsize=(16, 12))
for i, ax in enumerate(axes):
ax.imshow(img[:, :, i])
In [19]:
img = convert_color_rgb(test_image, 'HSV')
fig, axes = plt.subplots(1, 3, figsize=(16, 12))
for i, ax in enumerate(axes):
ax.imshow(img[:, :, i])
In [20]:
img = convert_color_rgb(test_image, 'HLS')
fig, axes = plt.subplots(1, 3, figsize=(16, 12))
for i, ax in enumerate(axes):
ax.imshow(img[:, :, i])
In [21]:
img = convert_color_rgb(test_image, 'YCrCb')
fig, axes = plt.subplots(1, 3, figsize=(16, 12))
for i, ax in enumerate(axes):
ax.imshow(img[:, :, i])
In [22]:
img = convert_color_rgb(test_image, 'LAB')
fig, axes = plt.subplots(1, 3, figsize=(16, 12))
for i, ax in enumerate(axes):
ax.imshow(img[:, :, i])
In [23]:
img_car_ch = img_car[:, :, 0]
features, hog_img = vis_hog_features(img_car_ch, orient=9, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
# Plot the examples
fig = plt.figure(figsize=(15, 10))
plt.subplot(231)
plt.imshow(img_car_ch)
plt.title('Channel 1')
plt.subplot(234)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
img_car_ch = img_car[:, :, 1]
features, hog_img = vis_hog_features(img_car_ch, orient=9, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
plt.subplot(232)
plt.imshow(img_car_ch)
plt.title('Channel 2')
plt.subplot(235)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
img_car_ch = img_car[:, :, 2]
features, hog_img = vis_hog_features(img_car_ch, orient=9, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
plt.subplot(233)
plt.imshow(img_car_ch)
plt.title('Channel 3')
plt.subplot(236)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
Out[23]:
In [24]:
img_car_cvt = cv2.cvtColor(img_car, cv2.COLOR_RGB2YUV)
img_car_ch = img_car[:, :, 0]
features, hog_img = vis_hog_features(img_car_ch, orient=9, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
# Plot the examples
fig = plt.figure(figsize=(15, 10))
plt.subplot(231)
plt.imshow(img_car_ch)
plt.title('Channel 1')
plt.subplot(234)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
img_car_ch = img_car_cvt[:, :, 1]
features, hog_img = vis_hog_features(img_car_ch, orient=9, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
plt.subplot(232)
plt.imshow(img_car_ch)
plt.title('Channel 2')
plt.subplot(235)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
img_car_ch = img_car_cvt[:, :, 2]
features, hog_img = vis_hog_features(img_car_ch, orient=9, pix_per_cell=16, cell_per_block=2, vis=True, feature_vec=False)
plt.subplot(233)
plt.imshow(img_car_ch)
plt.title('Channel 3')
plt.subplot(236)
plt.imshow(hog_img, cmap='gray')
plt.title('HOG Visualization')
Out[24]:
Features Classify¶
Three features are considered:
- Spatial bins
- histogram
- HOG
In [7]:
def read_image(img_path):
if img_path.endswith('.png'):
img = plt.imread(img_path)*255
img = np.clip(img, 0, 255).astype(np.uint8)
else:
img = plt.imread(img_path)
return img
def bin_spatial(img, size=(32, 32)):
features = cv2.resize(img, size).ravel()
return features
def color_hist(img, nbins=32): #bins_range=(0, 256)
# Compute the histogram of the color channels separately
channel1_hist = np.histogram(img[:,:,0], bins=nbins)
channel2_hist = np.histogram(img[:,:,1], bins=nbins)
channel3_hist = np.histogram(img[:,:,2], bins=nbins)
# Concatenate the histograms into a single feature vector
hist_features = np.concatenate((channel1_hist[0], channel2_hist[0], channel3_hist[0]))
# Return the individual histograms, bin_centers and feature vector
return hist_features
def convert_color_rgb(image, cspace='RGB'):
if cspace != 'RGB':
if cspace == 'HSV':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV)
elif cspace == 'LUV':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2LUV)
elif cspace == 'HLS':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2HLS)
elif cspace == 'YUV':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2YUV)
elif cspace == 'YCrCb':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2YCrCb)
elif cspace == 'LAB':
cvt_image = cv2.cvtColor(image, cv2.COLOR_RGB2LAB)
else:
cvt_image = np.copy(image)
return cvt_image
def extract_hog_features(img, orient, pix_per_cell, cell_per_block, feature_vec=True):
hog_features = hog(img, orientations=orient,
pixels_per_cell=(pix_per_cell, pix_per_cell),
cells_per_block=(cell_per_block, cell_per_block),
visualize=False, feature_vector=feature_vec, block_norm='L2-Hys')
return hog_features
def extract_features(image, cspace='RGB', orient=9, pix_per_cell=8,
spatial_size=(32, 32), hist_bins=32,
cell_per_block=2, hog_channel=0, spatial_feat=True,
hist_feat=True, hog_feat=True):
feature_image = convert_color_rgb(image, cspace)
img_features = []
if spatial_feat:
spatial_features = bin_spatial(feature_image, size=spatial_size)
img_features.append(spatial_features)
if hist_feat:
hist_features = color_hist(feature_image, nbins=hist_bins)
img_features.append(hist_features)
if hog_feat:
if hog_channel == 'ALL':
hog_features = []
for ch in range(feature_image.shape[2]):
hog_features_ch = extract_hog_features(image[:, :, ch], orient, pix_per_cell, cell_per_block, feature_vec=True)
hog_features.append(hog_features_ch)
hog_features = np.ravel(hog_features)
else:
hog_features = extract_hog_features(feature_image[:, :, hog_channel], orient,
pix_per_cell, cell_per_block, feature_vec=True)
img_features.append(hog_features)
return np.concatenate(img_features)
In [17]:
cspace = 'YUV' # Can be RGB, HSV, LUV, HLS, YUV, YCrCb
orient = 9
pix_per_cell = 8
cell_per_block = 2
hog_channel = 'ALL' # Can be 0, 1, 2, or "ALL"
spatial_size=(16, 16)
hist_bins = 8
spatial_feat = True
hist_feat = True
hog_feat = True
Read images with cars and without cars
In [18]:
#CAR_IMAGES_PATH = ROOT_PATH/'data/vehicles/KITTI_extracted'
# CAR_IMAGES2_PATH = ROOT_PATH/'data/vehicles/GTI_MiddleClose'
# CAR_IMAGES3_PATH = ROOT_PATH/'data/vehicles/GTI_Left'
# CAR_IMAGES4_PATH = ROOT_PATH/'data/vehicles/GTI_Right'
# CAR_IMAGES5_PATH = ROOT_PATH/'data/vehicles/GTI_Far'
CAR_IMAGES_PATH = ROOT_PATH/'data/vehicles/all'
NOTCAR_IMAGES_PATH = ROOT_PATH/'data/non-vehicles/GTI'
NOTCAR_IMAGES2_PATH = ROOT_PATH/'data/non-vehicles/Extras'
print('Total car images: ', len(list(CAR_IMAGES_PATH.iterdir())))
# print('Total car images: ', len(list(CAR_IMAGES2_PATH.iterdir())))
# print('Total car images: ', len(list(CAR_IMAGES3_PATH.iterdir())))
# print('Total car images: ', len(list(CAR_IMAGES4_PATH.iterdir())))
# print('Total car images: ', len(list(CAR_IMAGES5_PATH.iterdir())))
print('Not car iMages: ', len(list(NOTCAR_IMAGES_PATH.iterdir())))
print('Not car iMages: ', len(list(NOTCAR_IMAGES2_PATH.iterdir())))
In [19]:
%%time
SAMPLE_RATIO = 1
car_features = []
# car_files = (list(CAR_IMAGES_PATH.glob('*png')) +
# list(CAR_IMAGES2_PATH.glob('*png')) +
# list(CAR_IMAGES3_PATH.glob('*png')) +
# list(CAR_IMAGES4_PATH.glob('*png')) +
# list(CAR_IMAGES5_PATH.glob('*png')))
print('car features...')
car_files = list(CAR_IMAGES_PATH.glob('*png'))
num_files = len(car_files)
idx = np.random.choice(range(num_files), int(num_files*SAMPLE_RATIO))
for file in np.array(car_files)[idx]:
img = read_image(file.as_posix())
features = extract_features(img, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
car_features.append(features)
plt.imshow(img)
plt.show()
print('notcar features...')
notcar_features = []
noncar_files = list(NOTCAR_IMAGES_PATH.glob('*png')) + list(NOTCAR_IMAGES2_PATH.glob('*png'))
num_files = len(noncar_files)
idx = np.random.choice(range(num_files), int(num_files*SAMPLE_RATIO))
for file in np.array(noncar_files)[idx]:
img = read_image(file.as_posix())
features = extract_features(img, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
notcar_features.append(features)
plt.imshow(img)
plt.show()
In [20]:
X = np.vstack((car_features, notcar_features)).astype(np.float64)
print('Shape of X: ', X.shape)
In [21]:
y = np.hstack((np.ones(len(car_features)), np.zeros(len(notcar_features))))
print('Shape of y: ', y.shape)
Random split
In [22]:
rand_state = np.random.randint(0, 100)
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.2, random_state=rand_state)
In [23]:
# Fit a per-column scaler
X_scaler = StandardScaler().fit(X_train)
X_train = X_scaler.transform(X_train)
X_test = X_scaler.transform(X_test)
Training
In [24]:
print('Using:',orient,'orientations',pix_per_cell,
'pixels per cell and', cell_per_block,'cells per block')
print('Feature vector length:', len(X_train[0]))
# Use a linear SVC
svc = LinearSVC()
# Check the training time for the SVC
t=time.time()
svc.fit(X_train, y_train)
t2 = time.time()
print(round(t2-t, 2), 'Seconds to train SVC...')
# Check the score of the SVC
print('Test Accuracy of SVC = ', round(svc.score(X_test, y_test), 4))
# Check the prediction time for a single sample
t=time.time()
n_predict = 3000
#print('My SVC predicts: ', svc.predict(X_test[0:n_predict]))
#print('For these',n_predict, 'labels: ', y_test[0:n_predict])
t2 = time.time()
print(round(t2-t, 5), 'Seconds to predict', n_predict,'labels with SVC')
Sliding windows¶
In [25]:
def draw_boxes(img, bboxes, color=(0, 0, 255), thick=6):
# Make a copy of the image
imcopy = np.copy(img)
# Iterate through the bounding boxes
for bbox in bboxes:
# Draw a rectangle given bbox coordinates
cv2.rectangle(imcopy, bbox[0], bbox[1], color, thick)
# Return the image copy with boxes drawn
return imcopy
In [26]:
def slide_window(img, x_start_stop=[0, None], y_start_stop=[0, None],
xy_window=(64, 64), xy_overlap=(0.5, 0.5)):
if x_start_stop[1] == None:
x_start_stop[1] = img.shape[1]
if y_start_stop[1] == None:
y_start_stop[1] = img.shape[0]
xspan = x_start_stop[1] - x_start_stop[0]
yspan = y_start_stop[1] - y_start_stop[0]
nx_pix_per_step = np.int(xy_window[0]*(1-xy_overlap[0]))
ny_pix_per_step = np.int(xy_window[1]*(1-xy_overlap[1]))
nx_buffer = np.int(xy_window[0]*(xy_overlap[0]))
ny_buffer = np.int(xy_window[1]*(xy_overlap[1]))
# number of windows in x and y
nx_windows = np.int((xspan-nx_buffer)/nx_pix_per_step)
ny_windows = np.int((yspan-ny_buffer)/ny_pix_per_step)
window_list = []
for xs in range(nx_windows):
for ys in range(ny_windows):
startx = xs*nx_pix_per_step + x_start_stop[0]
endx = startx + xy_window[0]
starty = ys*ny_pix_per_step + y_start_stop[0]
endy = starty + xy_window[1]
window_list.append(((startx, starty), (endx, endy)))
return window_list
In [27]:
test_image_path = ROOT_PATH/'images/test_images/test5.jpg'
test_image = read_image(test_image_path.as_posix())
In [28]:
windows = slide_window(test_image, xy_window=(64, 64), xy_overlap=(0.5, 0.5))
img = draw_boxes(test_image, windows)
plt.figure(figsize=(12,9))
plt.imshow(img)
Out[28]:
Search and classify¶
In [29]:
# Define a function you will pass an image
# and the list of windows to be searched (output of slide_windows())
def search_windows(img, windows, clf, scaler, cspace='RGB',
spatial_size=(32, 32), hist_bins=32,
hist_range=(0, 256), orient=9,
pix_per_cell=8, cell_per_block=2,
hog_channel=0, spatial_feat=True,
hist_feat=True, hog_feat=True):
#1) Create an empty list to receive positive detection windows
on_windows = []
#2) Iterate over all windows in the list
for window in windows:
#3) Extract the test window from original image
test_img = cv2.resize(img[window[0][1]:window[1][1], window[0][0]:window[1][0]], (64, 64))
#4) Extract features for that window using single_img_features()
features = extract_features(test_img, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
#5) Scale extracted features to be fed to classifier
test_features = scaler.transform(np.array(features).reshape(1, -1))
#6) Predict using your classifier
prediction = clf.predict(test_features)
#7) If positive (prediction == 1) then save the window
if prediction == 1:
#print(features.shape)
#plt.imshow(test_img)
#plt.show()
on_windows.append(window)
#8) Return windows for positive detections
return on_windows
In [30]:
def slide_and_search(img, y_start_stop, classifier, scaler, xy_window, xy_overlap, cspace,
spatial_size, hist_bins, orient, pix_per_cell,
cell_per_block, hog_channel, spatial_feat=True, hist_feat=True, hog_feat=True):
# crop the image
windows = slide_window(img, xy_window=xy_window, xy_overlap=xy_overlap, y_start_stop=y_start_stop)
hot_windows = search_windows(img, windows, svc, X_scaler, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
return windows, hot_windows
In [31]:
def draw_windows(img, windows, hot_windows, vis_search=False):
if vis_search:
img = draw_boxes(img, windows, color=(242, 255, 191))
img = draw_boxes(img, hot_windows, color=(216, 48, 10))
return img
In [32]:
%%time
xy_window = (48, 48)
xy_overlap = (0.5, 0.5)
y_start_stop = (395, 500)
windows, hot_windows = slide_and_search(test_image, y_start_stop, svc, X_scaler, xy_window, xy_overlap, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
result_image = draw_windows(test_image, windows, hot_windows, vis_search=True)
plt.figure(figsize=(12, 9))
plt.imshow(result_image)
In [33]:
%%time
xy_window = (64, 64)
xy_overlap = (0.4, 0.4)
y_start_stop = (395, 550)
windows, hot_windows = slide_and_search(test_image, y_start_stop, svc, X_scaler, xy_window, xy_overlap, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
result_image = draw_windows(test_image, windows, hot_windows, vis_search=True)
plt.figure(figsize=(12, 9))
plt.imshow(result_image)
In [34]:
%%time
xy_window = (80, 80)
xy_overlap = (0.4, 0.4)
y_start_stop = (395, 560)
windows, hot_windows = slide_and_search(test_image, y_start_stop, svc, X_scaler, xy_window, xy_overlap, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
result_image = draw_windows(test_image, windows, hot_windows, vis_search=True)
plt.figure(figsize=(12, 9))
plt.imshow(result_image)
In [35]:
%%time
xy_window = (92, 92)
xy_overlap = (0.4, 0.4)
y_start_stop = (400, 600)
windows, hot_windows = slide_and_search(test_image, y_start_stop, svc, X_scaler, xy_window, xy_overlap, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
result_image = draw_windows(test_image, windows, hot_windows, vis_search=True)
plt.figure(figsize=(12, 9))
plt.imshow(result_image)
In [36]:
%%time
xy_window = (116, 116)
xy_overlap = (0.5, 0.5)
y_start_stop = (400, 700)
windows, hot_windows = slide_and_search(test_image, y_start_stop, svc, X_scaler, xy_window, xy_overlap, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
result_image = draw_windows(test_image, windows, hot_windows, vis_search=True)
plt.figure(figsize=(12, 9))
plt.imshow(result_image)
Multiple detections and false positive¶
In [37]:
def add_heat(heatmap, bbox_list):
# Iterate through list of bboxes
for box in bbox_list:
# Add += 1 for all pixels inside each bbox
# Assuming each "box" takes the form ((x1, y1), (x2, y2))
heatmap[box[0][1]:box[1][1], box[0][0]:box[1][0]] += 1
# Return updated heatmap
return heatmap# Iterate through list of bboxes
def apply_threshold(heatmap, threshold):
# Zero out pixels below the threshold
heatmap[heatmap <= threshold] = 0
# Return thresholded map
return heatmap
def draw_labeled_bboxes(img, labels):
# Iterate through all detected cars
for car_number in range(1, labels[1]+1):
# Find pixels with each car_number label value
nonzero = (labels[0] == car_number).nonzero()
# Identify x and y values of those pixels
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Define a bounding box based on min/max x and y
bbox = ((np.min(nonzerox), np.min(nonzeroy)), (np.max(nonzerox), np.max(nonzeroy)))
# Draw the box on the image
cv2.rectangle(img, bbox[0], bbox[1], (0,0,255), 6)
# Return the image
return img
# def filter_heatmap_with_thresh(heatmap, labels, thresh):
# num_labels = 0
# for car_number in range(1, labels[1]+1):
# nonzero = (labels[0] == car_number).nonzero()
# box_size = (np.max(nonzerox)-np.min(nonzerox))*(np.max(nonzeroy)-np.min(nonzeroy))
# if box_size < thresh:
# labels[0][nonzeros[0], nonzero[1]] = 0
In [38]:
%%time
test_image_path = ROOT_PATH/'images/test_images/test6.jpg'
test_image = read_image(test_image_path.as_posix())
window_list = [
(48, 0.5, 395, 500),
(64, 0.5, 405, 520),
(80, 0.5, 395, 560),
#(92, 0.5, 400, 600),
(116, 0.5, 400, 700)
]
threshold_detect = 3
def detect_cars(img, window_list, threshold_detect, classifier,
scaler, cspace, spatial_size, hist_bins, orient, pix_per_cell,
cell_per_block, hog_channel, spatial_feat, hist_feat, hog_feat):
heat = np.zeros_like(test_image[:,:,0]).astype(np.float)
for window, overlap, ystart, ystop in window_list:
xy_window = (window, window)
xy_overlap = (overlap, overlap)
y_start_stop = (ystart, ystop)
_, hot_windows = slide_and_search(img, y_start_stop, classifier, scaler, xy_window, xy_overlap,
cspace=cspace, spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell, cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
heat = add_heat(heat, hot_windows)
# apply the threhold
heat = apply_threshold(heat, threshold_detect)
# create heatmap and labels
heatmap = np.clip(heat, 0, 255)
labels = label(heatmap)
return heatmap, labels
heatmap, labels = detect_cars(test_image, window_list, threshold_detect, svc, X_scaler, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins, orient=orient,
pix_per_cell=pix_per_cell, cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat, hist_feat=hist_feat,
hog_feat=hog_feat)
In [39]:
for img_path in (ROOT_PATH/'images/test_images').glob('*jpg'):
print(img_path)
img = read_image(img_path.as_posix())
heatmap, labels = detect_cars(img, window_list, threshold_detect, svc, X_scaler, cspace=cspace,
spatial_size=spatial_size, hist_bins=hist_bins, orient=orient,
pix_per_cell=pix_per_cell, cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat, hist_feat=hist_feat,
hog_feat=hog_feat)
print('number of labels: ', labels[1])
draw_img = draw_labeled_bboxes(np.copy(img), labels)
fig = plt.figure(figsize=(16,12))
plt.subplot(121)
plt.imshow(draw_img)
plt.title('Car Positions')
plt.subplot(122)
plt.imshow(heatmap, cmap='hot')
plt.title('Heat Map')
fig.tight_layout()
plt.show()
HOG Subsampling¶
In [40]:
# Define a single function that can extract features using hog sub-sampling and make predictions
def find_cars(img, cspace, cells_per_step, ystart, ystop, scale, svc, X_scaler,
orient, pix_per_cell, cell_per_block, spatial_size, hist_bins):
draw_img = np.copy(img)
#img = img.astype(np.float32)/255
img_tosearch = img[ystart:ystop,:,:]
ctrans_tosearch = convert_color_rgb(img_tosearch, cspace=cspace)
if scale != 1:
imshape = ctrans_tosearch.shape
ctrans_tosearch = cv2.resize(ctrans_tosearch, (np.int(imshape[1]/scale), np.int(imshape[0]/scale)))
print('ctrans_tosearch shape: ', ctrans_tosearch.shape)
ch1 = ctrans_tosearch[:,:,0]
ch2 = ctrans_tosearch[:,:,1]
ch3 = ctrans_tosearch[:,:,2]
#ch1 = ctrans_tosearch[:,:,0]
# Define blocks and steps as above
nxblocks = (ch1.shape[1] // pix_per_cell) - cell_per_block + 1
nyblocks = (ch1.shape[0] // pix_per_cell) - cell_per_block + 1
#nfeat_per_block = orient*cell_per_block**2
# 64 was the orginal sampling rate, with 8 cells and 8 pix per cell
window = 64
nblocks_per_window = (window // pix_per_cell) - cell_per_block + 1
cells_per_step = cells_per_step # Instead of overlap, define how many cells to step
nxsteps = (nxblocks - nblocks_per_window) // cells_per_step + 1
nysteps = (nyblocks - nblocks_per_window) // cells_per_step + 1
# Compute individual channel HOG features for the entire image
hog1 = extract_hog_features(ch1, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog2 = extract_hog_features(ch2, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog3 = extract_hog_features(ch3, orient, pix_per_cell, cell_per_block, feature_vec=False)
#hog = extract_hog_features(ch1, orient, pix_per_cell, cell_per_block, feature_vec=False)
for xb in range(nxsteps):
for yb in range(nysteps):
ypos = yb*cells_per_step
xpos = xb*cells_per_step
# Extract HOG for this patch
hog_feat1 = hog1[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat2 = hog2[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat3 = hog3[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_features = np.hstack((hog_feat1, hog_feat2, hog_feat3))
#hog_features = hog[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
xleft = xpos*pix_per_cell
ytop = ypos*pix_per_cell
# Extract the image patch
subimg = cv2.resize(ctrans_tosearch[ytop:ytop+window, xleft:xleft+window], (64, 64))
# Get color features
spatial_features = bin_spatial(subimg, size=spatial_size)
hist_features = color_hist(subimg, nbins=hist_bins)
# Scale features and make a prediction
test_features = X_scaler.transform(np.hstack((spatial_features, hist_features, hog_features)).reshape(1, -1))
test_prediction = svc.predict(test_features)
if test_prediction == 1:
xbox_left = np.int(xleft*scale)
ytop_draw = np.int(ytop*scale)
win_draw = np.int(window*scale)
cv2.rectangle(draw_img,(xbox_left, ytop_draw+ystart),
(xbox_left+win_draw,ytop_draw+win_draw+ystart),(0,0,255),6)
return draw_img
In [197]:
%%time
test_image_path = ROOT_PATH/'images/test_images/test3.jpg'
test_image = read_image(test_image_path.as_posix())
ystart = 402
ystop = 500
scale = 0.98
cells_per_step = 4
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
In [198]:
%%time
ystart = 390
ystop = 550
scale = 1
cells_per_step = 4
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
In [202]:
%%time
ystart = 400
ystop = 550
scale = 1.25
cells_per_step = 4
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
In [187]:
%%time
ystart = 390
ystop = 600
scale = 1.5
cells_per_step = 6
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
In [188]:
%%time
ystart = 395
ystop = 600
scale = 1.75
cells_per_step = 6
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
In [189]:
%%time
ystart = 390
ystop = 650
scale = 2.25
cells_per_step = 4
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
In [208]:
%%time
ystart = 395
ystop = 660
scale = 2.5
cells_per_step = 2
out_img = find_cars(test_image, cspace, cells_per_step, ystart, ystop, scale, svc,
X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
plt.figure(figsize=(12, 9))
plt.imshow(out_img)
Subsampling with multiple windows¶
In [157]:
def search_windows_with_scale(img, scale, cells_per_step, ystart, ystop, classifier,
scaler, cspace, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins):
hot_windows = []
draw_img = np.copy(img)
# crate an empty list to receive positive detection windows
img_tosearch = img[ystart:ystop,:,:]
ctrans_tosearch = convert_color_rgb(img_tosearch, cspace=cspace)
if scale != 1:
imshape = ctrans_tosearch.shape
ctrans_tosearch = cv2.resize(ctrans_tosearch, (np.int(imshape[1]/scale), np.int(imshape[0]/scale)))
#print('ctrans_tosearch shape: ', ctrans_tosearch.shape)
ch1 = ctrans_tosearch[:,:,0]
ch2 = ctrans_tosearch[:,:,1]
ch3 = ctrans_tosearch[:,:,2]
# Define blocks and steps as above
nxblocks = (ch1.shape[1] // pix_per_cell) - cell_per_block + 1
nyblocks = (ch1.shape[0] // pix_per_cell) - cell_per_block + 1
#nfeat_per_block = orient*cell_per_block**2
# 64 was the orginal sampling rate, with 8 cells and 8 pix per cell
window_size = 64
nblocks_per_window = (window_size // pix_per_cell) - cell_per_block + 1
cells_per_step = cells_per_step # Instead of overlap, define how many cells to step
nxsteps = (nxblocks - nblocks_per_window) // cells_per_step + 1
nysteps = (nyblocks - nblocks_per_window) // cells_per_step + 1
# Compute individual channel HOG features for the entire image
#hog = extract_hog_features(ch1, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog1 = extract_hog_features(ch1, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog2 = extract_hog_features(ch2, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog3 = extract_hog_features(ch3, orient, pix_per_cell, cell_per_block, feature_vec=False)
for xb in range(nxsteps):
for yb in range(nysteps):
ypos = yb*cells_per_step
xpos = xb*cells_per_step
# Extract HOG for this patch
#hog_features = hog[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat1 = hog1[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat2 = hog2[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat3 = hog3[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_features = np.hstack((hog_feat1, hog_feat2, hog_feat3))
xleft = xpos*pix_per_cell
ytop = ypos*pix_per_cell
# Extract the image patch
subimg = cv2.resize(ctrans_tosearch[ytop:ytop+window_size, xleft:xleft+window_size], (64, 64))
# Get color features
spatial_features = bin_spatial(subimg, size=spatial_size)
hist_features = color_hist(subimg, nbins=hist_bins)
# Scale features and make a prediction
#test_features = scaler.transform(np.hstack((spatial_features, hog_features)).reshape(1, -1))
test_features = X_scaler.transform(np.hstack((spatial_features, hist_features, hog_features)).reshape(1, -1))
test_prediction = classifier.predict(test_features)
if test_prediction == 1:
xbox_left = np.int(xleft*scale)
ytop_draw = np.int(ytop*scale)
win_draw = np.int(window_size*scale)
hot_window = ((xbox_left, ytop_draw+ystart), (xbox_left+win_draw, ytop_draw+win_draw+ystart))
hot_windows.append(hot_window)
return hot_windows
In [209]:
scales = [
#(0.9, 2, 410, 500),
(1, 4, 390, 550),
(1.25, 4, 400, 550),
(1.75, 6, 395, 650),
(2.25, 4, 390, 650),
(2.5, 2, 395, 660)
]
threshold = 3
In [210]:
def detect_with_scales(img, scales, threshold, classifier, scaler, cspace, orient,
pix_per_cell, cell_per_block, spatial_size, hist_bins):
# create an empty array to store heatmaps
heat = np.zeros_like(img[:,:,0]).astype(np.float)
for scale, cells_per_step, ystart, ystop in scales:
hot_windows = search_windows_with_scale(img, scale, cells_per_step, ystart, ystop, svc,
scaler, cspace, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
heat = add_heat(heat, hot_windows)
# apply the threshold
heatmap = apply_threshold(heat, threshold)
labels = label(heatmap)
return heatmap, labels
In [211]:
%%time
test_image_path = ROOT_PATH/'images/test_images/test3.jpg'
test_image = read_image(test_image_path.as_posix())
heatmap, labels = detect_with_scales(test_image, scales, threshold, svc, X_scaler, cspace,
orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
In [212]:
for img_path in (ROOT_PATH/'images/test_images').glob('*jpg'):
print(img_path)
img = read_image(img_path.as_posix())
heatmap, labels = detect_with_scales(img, scales, threshold, svc, X_scaler, cspace,
orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
print('number of labels: ', labels[1])
draw_img = draw_labeled_bboxes(np.copy(img), labels)
fig = plt.figure(figsize=(16,12))
plt.subplot(121)
plt.imshow(draw_img)
plt.title('Car Positions')
plt.subplot(122)
plt.imshow(heatmap, cmap='hot')
plt.title('Heat Map')
fig.tight_layout()
plt.show()
